Most CRM and CDP projects I work on in Germany don’t fail because of the technology. They fail before the first data pipeline is built. CMOs still largely lead marketing strategy within silos, brand A, channel B, campaign C. But the customer doesn’t see silos. They are a business traveller booking an upgrade on Monday. A family buyer hunting a weekend sale on Friday. An insurance policyholder renewing in October. The same person. Three distinct micropersonas. Three different touchpoints. And in the database: a flat profile with a date of birth and a list of all transactions.
When a CDP is placed on top of this foundation, without a clear strategy, without data governance, without defined ownership of matching logic and profile semantics, you don’t get a unified customer view. You get expensive, well-marketed data noise.
This is not theoretical. I see it regularly in German companies: customer databases with 40% duplicate records. Behavioural profiles built on click data stripped of context. AI models trained on dirty data that produce precisely wrong recommendations. And at the end: a personalisation initiative that gets quietly shut down, not because AI doesn’t work, but because the foundation was never laid.
It is no coincidence that a new role is emerging: the Chief AI & Digital Products Officer, hired specifically to close the gap that opened up between CMO, CRM, and IT.
The noise is getting louder, and German patience is running out
According to Gartner (April 2026), 81% of consumers actively try to tune out advertising. In Germany, that is not surprising. Around 40% of desktop users run ad blockers (Statista, 2020). YouTube Premium, Spotify Premium, ad-free subscriptions to Spiegel or Die Zeit, people are paying to escape. Industry estimates put the daily volume of brand messages in the thousands; the precise number is hard to verify, but the direction is clear.
And the paid media landscape keeps expanding. ChatGPT launched advertising formats in February 2026, reportedly generating around $100M in annualised revenue at a $60 CPM. Google is inserting ads into 25.5% of AI-generated answers. AI search is the next paid channel, with higher CPMs and still-unclear performance metrics. Perplexity scrapped its ad model in February 2026. More channels. More noise. Higher cost. Less tolerance.
The conclusion is structural: direct, consent-based customer relationships are not just strategically more valuable; they are the only channel that cannot be devalued by algorithm changes or platform policy shifts.
Agentic AI amplifies the problem or the solution
Every major platform now promises AI agents: Salesforce Agentforce, BrazeAI with Operator and Decisioning Studio, MoEngage with the Sherpa AI engine (Intelligent Path Optimizer, Merlin AI, Proactive Assistant). The promises are real. The prerequisites are almost always underestimated.
An AI agent running on bad data is not progress. It is a scaled-up version of the mistake. It sends the wrong message to more people, faster, more automatically. And in Germany, where consumers are already deeply sceptical, in our “Mind the AI Gap” study (Artefact × MoEngage, 1,000+ consumers across the UK, Germany, and the Netherlands), 34% answered “None of these” when asked what excites them about AI-driven engagement; that is not an abstract risk. 41% of respondents identify transparent and ethical data practices as the most critical AI capability for brand loyalty. Not personalisation quality. Not channel breadth. Trust in how data is handled. 29% feel constantly sold to rather than genuinely helped. 39% want AI that clearly benefits them, not just the business.
German consumers don’t forgive relevance failures. They unsubscribe.
What works and in what order
The failure I see most often is not about willingness. It is about sequence. The right order looks something like this(not in linear order, one can do strategy & data governance setup in parallel):
Strategy first. Before a CDP is selected or an engagement tool configured, these questions must be answered: What is our definition of a customer, across all brands, channels, and product lines? What micropersonas exist, and which journey moments are decisive? Who in the organisation owns these definitions? Without these answers, any technology project is capital deployed on poor foundations.
Data governance second. Customer profile matching logic, consent architecture under GDPR, data ownership rules, quality thresholds for AI training- these are not IT decisions. They are business decisions that must be made from the top, before the first data engineer starts work.
Architecture third. The key question: do you need a standalone CDP, or does a modern data warehouse architecture with reverse-ETL tools like Hightouch solve the problem more efficiently? Leading customer engagement platforms such as Braze, MoEngage, Bloomreach, and Insider are increasingly building CDP functionality in-house. For many mid-sized German companies, that is sufficient, provided the data foundation is solid.
AI and agents last. Only on this basis does agentic AI become what the platform promises to describe: a system that autonomously optimises journeys, computes next-best-action in real time, and generates channel-specific content — with real customer knowledge rather than statistical noise.
A warning on tool selection and RFPs
Not all CEPs, CDPs, and CRM systems are equal, and that is especially true by industry. A CEP built for D2C e-commerce does not fit complex B2B2C journeys in insurance or telecoms. A CDP with strong retail connectors can create fundamental matching problems in a travel or FMCG architecture.
The most common mistake I see in German RFP processes: companies evaluate features instead of fit. They compare function lists, not use-case coverage for their specific journey complexity and data situation. The result: a platform that shines in the demo and fails in production because the data model does not fit, because the team lacks the ownership structures the tool assumes, or because pricing add-ons (AI modules, channel volumes, API calls) cut the business case in half by year two.
MarTech RFP processes are demanding. But those who shortcut them pay multiples later.
What happens if you don’t get this right
This is the question missing from most board presentations.
Compliance. The EU AI Act and Germany’s KI-MIG (in force since February 2026) tighten requirements for AI-driven profiling and automated decision-making. Consent architectures that were already thin under GDPR become a compliance liability under these frameworks. Fines are not hypothetical; they are in the pipeline for companies with demonstrably poor data practices.
Brand trust. When an AI agent triggers an irrelevant, mistimed, or wrong-channel message in Germany, where 34% of consumers already answer “None of these” to AI engagement, where WhatsApp is treated as a private channel and push notifications as a tolerance rather than a welcome, opt-out rates are high, and return rates are low. Trust burned by bad AI is slow to rebuild.
Competitive loss. Brands that invest now in clean first-party data, consent-first personalisation, and a solid CDP architecture will build a structural advantage over the next 24 months that no paid media budget can buy back. RCS already reaches 88% smartphone penetration in Germany, a channel that opens a new quality of direct communication for brands with clean opt-in. Those without the opt-in cannot use the channel, regardless of how good their technology stack is.
The bottom line
CRM is not an email tool. CDP is not the database project. And agentic AI is not an automation upgrade; it is a multiplier. A multiplier on what is already there. Organisations with bad data, missing governance, and fragmented journey strategy will scale exactly that.
The question for CMOs and CRM leaders in Germany is not: “Which platform should we buy?” It is: “Do we have the strategic clarity, the data governance, and the organisational ownership that these platforms assume?”
Those who answer that question honestly with “no” and then buy the technology anyway should not be surprised when a Chief AI & Digital Products Officer is hired two years later to sort through the wreckage.
